Competing Bioaerosols May Influence the Seasonality of Influenza-Like Illnesses, including COVID-19. The Chicago Experience

Data from Chicago confirm the end of flu season coincides with the beginning of pollen season. More importantly, the end of flu season also coincides with onset of seasonal aerosolization of mold spores. Overall, the data suggest bioaerosols, especially mold spores, compete with viruses for a shared receptor, with the periodicity of influenza-like illnesses, including COVID-19, a consequence of seasonal factors that influence aerosolization of competing species.


Introduction
Influenza-like illnesses (ILIs) attributable to influenza viruses and to coronaviruses are sharply seasonal [1,2]. Importantly, recent data from the Netherlands indicate there exists an inverse relationship between the seasonal incidence of ILIs, including COVID-19, and pollen count [3,4]. To discern whether such a relationship might be the case generally, pollen count in Chicago was related to ILIs reported by local emergency departments. In Chicago, as in the Netherlands, ILIs fall as total pollen count rises.
Because bioaerosols measured in Chicago include not only pollens but also mold spores, ILIs were related to counts of the two measured species. Just as they do for pollens, ILIs fall as mold spores rise. In contradistinction to their temporal relationship with pollens, however, ILIs remain low when mold spores are high, rising again when mold spores fall.
Perusal of the various measured pollens and mold spores reveals many are echinulated, having protuberances reminiscent of those of influenza and SARS-CoV-2 virions [5][6][7][8][9][10][11]. In addition, although interactions of viral 'spikes' with other host proteins have been documented [12,13], such protuberances seem ideally suited to interacting especially with Toll-like receptors in a fashion akin to 'hook-and-loop' adhesives. Indeed, such interactions are well described for a number of pollens and mold spores, and there is no reason to believe viruses, despite their smaller size, interact differently. Implicating Toll-like receptors, Tolllike receptor 4 (TLR4) in particular, seems appropriate on phenomenological grounds as well: (1) engagement of TLR 4 can account for the inflammatory signaling characteristic of severe COVID-19 [14,15], (2) TLR4 is intimately involved in the inflammation elicited both by sharply seasonal respiratory viruses and by multiple species of fungi [16][17][18][19][20],

ILI and COVID-19 Data
ILI data pooled from 23 large hospitals in Chicago over the period 9 January 2015 through 18 July 2020 were obtained from the Chicago Department of Public Health (CDPH). The 23 hospitals were chosen because they alone of Chicago-area hospitals consistently reported ILI presentations over the entirety of the study interval. In Chicago, the designation of ILI by emergency departments is based on fever (≥100 • F) and respiratory symptoms, i.e., cough and/or sore throat, not on any specific diagnosis. The data are  Supplementary Table S3. COVID-19 data from all Chicago hospitals were obtained through portals of the CDPH, https://www.chicago.gov/city/en/sites/covid19/ home/covid-dashboard.html (accessed on 18 July 2020) and https://data.cityofchicago. org/browse?limitTo=datasets&sortBy=alpha&tags=covid-19 (accessed on 18 November 2020). Those data are also included in Supplementary Table S4.

Data Analysis
Time-or dose-dependent data were paired with the corresponding time or dose and fit to equations described in the text. The best values for the parameters of the equations, as well as their corresponding 95% confidence intervals, were then determined using the paired data and a nonlinear least squares regression method [29].

Kinetics of Presentations to Emergency Departments
As shown in Figure 1A, ILI presentations to emergency departments in Chicago are cyclical in nature, with a periodicity of~1 year. The peak incidence occurs approximately during February, with the annual nadir occurring approximately during August. Over the entirety of a seasonal cycle, ILI presentations never reach zero. Although there are subtleties associated with the kinetics of ILIs for each of the individual years 2015 through 2020, the annual increase in ILI presentations for each season is characterized by a leading 'bump,' followed thereafter by a major rise. The relevance of the bump is addressed below, under both Results and Discussion. As for the major rise, it is approximated empirically by a first-order process, ILI obs = ILI 0 e kt + C, where ILI obs is the number of observed ILIs during the growth phase of a cycle; ILI 0 is the number of ILI presentations at the start of a cycle; t is time, in d; k is a first-order rate constant, in d −1 ; and C is the number of background presentations 'masquerading' as ILIs (Table 2 and Figure 2A). The rate of decrease from any yearly maximum also fits reasonably well a first-order process (Table 2 and Figure 2B). Taken at face value, the data suggest: (1) a not insignificant number of ILI presentations are due to pathogens other than influenza virus, e.g., parainfluenza virus, respiratory syncytial virus, measles, mumps, etc., and (2) the annual rate of change in influenza cases is due to change in ambient concentration of influenza virus.      Figure 1A) to a first-order rate equation, the parameters of which are given in Table 2. The inset shows the expected linearity of the same data when plotted according to the method of Kézdy [30], in this case ILI presentationsday n vs. ILI presentationsday n+1.  Figure 1A) to a first-order rate equation, the parameters of which are given in Table 2. The inset shows the expected linearity of the same data when plotted according to the method of Kézdy [30], in this case ILI presentations day n vs. ILI presentations day n+1 . Figure 3A shows the time course of the 7-day moving average of COVID-19 presentations to emergency departments of all Chicago hospitals. Because reporting was not uniformly rigorous before 1 May 2020, COVID-19 presentations prior to that date have been excluded from analyses. As shown in Figure 3B, the fall in presentations for the period 5 May 2020 through 27 September 2020 was roughly first-order, with parameters k = 0.065 d −1 (t 1/2~1 0.6 d), COVID-19 0 = 875 presentations and C = 262 presentations. Inasmuch as masks and physical-distancing were mandated in Chicago on 1 May 2020, the rate of fall in COVID-19 presentations, i.e., the shape of the curve, after that date was influenced to some extent by those measures, as discussed elsewhere [31]. Figure 3A shows the time course of the 7-day moving average of COVID-19 presentations to emergency departments of all Chicago hospitals. Because reporting was not uniformly rigorous before 1 May 2020, COVID-19 presentations prior to that date have been excluded from analyses. As shown in Figure 3B, the fall in presentations for the period 5 May 2020 through 27 September 2020 was roughly first-order, with parameters k = 0.065 d −1 (t1/2 ~ 10.6 d), COVID-190 = 875 presentations and C = 262 presentations. Inasmuch as masks and physical-distancing were mandated in Chicago on 1 May 2020, the rate of fall in COVID-19 presentations, i.e., the shape of the curve, after that date was influenced to some extent by those measures, as discussed elsewhere [31].
Starting in late September 2020, COVID-19 cases in Chicago surged in first-order fashion ( Figure 3C), the parameters of which are k = 0.053 d −1 (t1/2 ~ 13 d), COVID-190 = 96 presentations and C = 117 presentations. Taken at face value, the data suggest: (1) some of the individuals for whom a diagnosis of COVID-19 was made did not have COVID-19 and (2) changes in the number of COVID-19 presentations in Chicago were due to changes in the ambient concentration of SARS-CoV-2.   13 September 2020. The dashed line displayed in the primary plot is the theoretical fit of the data to a first-order rate equation, the parameters of which are given in the text. For more refined analysis, see [31]. (C) Appearance of COVID-19 presentations, 14 September 2020 to 14 November 2020. The dashed line displayed in the primary plot is the theoretical fit of the data to a first-order rate equation, the parameters of which are given in the text. The insets show expected linearities of the data when plotted according to the method of Kézdy [30], in these cases COVID-19 presentations day n vs. COVID-19 presentations day n+1 .
Starting in late September 2020, COVID-19 cases in Chicago surged in first-order fashion ( Figure 3C), the parameters of which are k = 0.053 d −1 (t 1/2~1 3 d), COVID-19 0 = 96 presentations and C = 117 presentations. Taken at face value, the data suggest: (1) some of the individuals for whom a diagnosis of COVID-19 was made did not have COVID-19 and (2) changes in the number of COVID-19 presentations in Chicago were due to changes in the ambient concentration of SARS-CoV-2.

Kinetics of Pollen and Mold Spore Counts
Pollens are fertilizing elements of flowering plants whilst mold spores are reproductive elements of fungi. In published studies [3,4], pollens alone were counted and related to ILIs. Left uncounted were mold spores, important seasonal contributors to the total bioaerosol burden. For the studies reported herein, both pollens and mold spores were counted. Those counts were then analyzed, in aggregate and individually.
In Chicago, pollens and mold spores are monitored from approximately mid-March to approximately mid-October, the only time during which the bioaerosols are easily measurable and also the time most problematic for persons suffering from seasonal allergies. As expected, the data indicate bioaerosol expression is cyclical with a periodicity of~1 year ( Figure 1B). The total bioaerosol count peaks during approximately mid-September and falls sharply thereafter. Data following the peaks are somewhat limited, their collection being truncated on an arbitrary end date, i.e., approximately mid-October.
In the case of pollens, the seasonal distribution is bimodal, with a dominant first mode that peaks in approximately mid-May and a smaller second mode that peaks in approximately late August ( Figure 1C). The pollens that constitute the second mode, here termed 'late pollens', are predominantly Ambrosia and the other Asteraceae. Importantly, the peak of the second mode always coincides with the leading bump in ILI presentations ( Figures 1A and 4). The potential relevance of this is addressed in the Discussion. In the case of mold spores, which constitute the bulk of the measured bioaerosols (Figure 1B,D  and Table 3), the peak count, which occurs during approximately late September, falls precipitously by approximately mid-October, with an empiric half-life of~10 d, Table 4. given in the text. The insets show expected linearities of the data when plotted according to the method of Kézdy [30], in these cases COVID-19 presentationsday n vs. COVID-19 presentationsday n+1.

Kinetics of Pollen and Mold Spore Counts
Pollens are fertilizing elements of flowering plants whilst mold spores are reproductive elements of fungi. In published studies [3,4], pollens alone were counted and related to ILIs. Left uncounted were mold spores, important seasonal contributors to the total bioaerosol burden. For the studies reported herein, both pollens and mold spores were counted. Those counts were then analyzed, in aggregate and individually.
In Chicago, pollens and mold spores are monitored from approximately mid-March to approximately mid-October, the only time during which the bioaerosols are easily measurable and also the time most problematic for persons suffering from seasonal allergies. As expected, the data indicate bioaerosol expression is cyclical with a periodicity of ~1 year ( Figure 1B). The total bioaerosol count peaks during approximately mid-September and falls sharply thereafter. Data following the peaks are somewhat limited, their collection being truncated on an arbitrary end date, i.e., approximately mid-October.
In the case of pollens, the seasonal distribution is bimodal, with a dominant first mode that peaks in approximately mid-May and a smaller second mode that peaks in approximately late August ( Figure 1C). The pollens that constitute the second mode, here termed 'late pollens', are predominantly Ambrosia and the other Asteraceae. Importantly, the peak of the second mode always coincides with the leading bump in ILI presentations ( Figures 1A and 4). The potential relevance of this is addressed in the Discussion. In the case of mold spores, which constitute the bulk of the measured bioaerosols ( Figures 1B and 1D and Table 3), the peak count, which occurs during approximately late September, falls precipitously by approximately mid-October, with an empiric half-life of ~10 d, Table 4.  Although these data substantiate the claim of an inverse relationship between the onset of pollen season and the end of flu season, pollen count declines rapidly and is not elevated when ILI presentations ( Figure 5A) and COVID-19 presentations ( Figure 6A) are low. Mold spores, on the other hand, increase continuously in first-order fashion (Table 4 and Figure 7), beginning just prior to or coincident with the fall in ILI ( Figure 5B) and  (Figure 6B), and across the entirety of the summer months, when influenza and COVID-19 cases are low.  Although these data substantiate the claim of an inverse relationship between the onset of pollen season and the end of flu season, pollen count declines rapidly and is not elevated when ILI presentations ( Figure 5A) and COVID-19 presentations ( Figure 6A) are low. Mold spores, on the other hand, increase continuously in first-order fashion (Table 4 and Figure 7), beginning just prior to or coincident with the fall in ILI ( Figure 5B) and COVID-19 presentations (Figure 6B), and across the entirety of the summer months, when influenza and COVID-19 cases are low.  Figure 4 for additional details. The onset of aerosolization of mold spores also coincides with the drop in seasonal ILI presentations. Thereafter, mold spore count increases across the entirety of the summer months-during which time ILIs remain low-and falls precipitously from mid-September to mid-October, at which time ILI cases begin to rise. See text for additional details.   Table 4. The inset shows the expected linearity of the same data when plotted according to the method of Kézdy [30], in this case mold spore countday n vs. mold spore countday n+1.

Inhibition of ILI and COVID-19 Presentations by Mold Spores
If one assumes ILI and COVID-19 presentations are consequences of the binding of relevant viruses to specific receptors, then one can treat the presentations as proxies for those receptors, for which mold spores compete. Toward that end, ILI and COVID-19 presentations were plotted as functions of total mold spore count (Figure 8). Because the curvatures of the plots suggest true equilibria, the data of each were fit to the equation P = Po/(1 + C/Kd) + B, where P is the observed number of presentations to emergency departments; Po is the maximum number of such presentations; C, in mold spores/m 3 , is the measured mold spore count; Kd, in mold spores/m 3 , is the apparent dissociation constant of the receptor-mold spore complex; and B is a constant representing presentations not influenced by mold spores. As shown in the figures, the data of each plot fit the theoretical model reasonably well. From the ILI data, one calculates Po ~ 50 presentations, Kd ~ 2128 mold spores/m 3 and B ~ 16 presentations; from the COVID-19 data, one calculates Po ~ 1366 presentations, Kd ~ 1668 mold spores/m 3 and B ~ 201 presentations. The most parsimonious explanation for the near equivalence of the apparent dissociation constants is a shared receptor.  Table 4. The inset shows the expected linearity of the same data when plotted according to the method of Kézdy [30], in this case mold spore count day n vs. mold spore count day n+1 .

Inhibition of ILI and COVID-19 Presentations by Mold Spores
If one assumes ILI and COVID-19 presentations are consequences of the binding of relevant viruses to specific receptors, then one can treat the presentations as proxies for those receptors, for which mold spores compete. Toward that end, ILI and COVID-19 presentations were plotted as functions of total mold spore count (Figure 8). Because the curvatures of the plots suggest true equilibria, the data of each were fit to the equation P = P o /(1 + C/K d ) + B, where P is the observed number of presentations to emergency departments; P o is the maximum number of such presentations; C, in mold spores/m 3 , is the measured mold spore count; K d , in mold spores/m 3 , is the apparent dissociation constant of the receptor-mold spore complex; and B is a constant representing presentations not influenced by mold spores. As shown in the figures, the data of each plot fit the theoretical model reasonably well. From the ILI data, one calculates P o~5 0 presentations, K d~2 128 mold spores/m 3 and B~16 presentations; from the COVID-19 data, one calculates P o~1 366 presentations, K d~1 668 mold spores/m 3 and B~201 presentations. The most parsimonious explanation for the near equivalence of the apparent dissociation constants is a shared receptor.

Discussion
The data presented herein are consistent with those presented earlier by others [3,4], namely, the incidence of ILIs falls as pollen count rises. Because the data of the present study derive from an urban area in North America (Chicago, IL, USA: latitude 41.85003,

Discussion
The data presented herein are consistent with those presented earlier by others [3,4], namely, the incidence of ILIs falls as pollen count rises. Because the data of the present study derive from an urban area in North America (Chicago, IL, USA: latitude 41.85003, longitude −87.65005) whilst those of the earlier study derive from North Central Europe (Helmond, the Netherlands: latitude 51.48167, longitude 5.66111), it appears the inverse relationship may be generally valid.
Made blatantly obvious by these studies are the seasonalities of ILIs, pollens and mold spores. The annual periodicities of the three indicate the rotation of the earth about the sun is ultimately responsible. Special note should be made of: (1) the nearness of the onset of bioaerosol expression to the vernal equinox, i.e., when the lengths of day and night are nearly equal, (2) the nearness of the peak in pollen count (excepting Ambrosia) to the summer solstice, i.e., the longest day of the year and (3) the nearness of the peak in mold spore count to the autumnal equinox. Because light and heat from the sun are drivers of both natural and agricultural growing seasons, these dates and their relevance to the expression and dispersal of bioaerosols should come as no surprise.
With special regard to the late pollens, changes in their atmospheric concentration invariably coincide with the annual leading bump in ILIs. Inasmuch as Ambrosia, the dominant species, is a major respiratory allergen, the leading bump may represent ragweed sensitivities manifesting as ILI. Alternatively, it may just represent enhanced spread of respiratory viruses by, for example, school openings. Regardless, the peak in late pollen count-as if a switch-presages the major upswing in ILI ( Figure 4) and COVID-19 ( Figure 6C) presentations. Thus, aside from any contribution to mechanistic understanding it might provide, the peak in late pollens could be exploited when contemplating an upcoming ILI season.
From an anthropologic perspective, the potential of mold spores and pollens to inhibit influenza-like epidemics/pandemics, including COVID-19, certainly has great relevance and significant consequence. Still, because in comparison to plants, fungi and even viruses, humans contribute only very little to the biomass on planet Earth [32], it seems likely some larger purpose is served by interplay between the three bioaerosols. It is tempting to speculate that any antiviral effect attributable to mold spores and/or pollens is intended to benefit primarily fungi and plants [33], i.e., the human benefit, albeit perhaps related mechanistically, is an epiphenomenon. As just one of many possibilities, mold spores and pollens might protect primarily arthropods, birds and bats, organisms intimately involved in dissemination of reproductive elements of both fungi and plants [34][35][36][37][38][39][40][41].
Separate and distinct from pollen count, mold spore count in Chicago correlates inversely with ILIs. Indeed, given their higher atmospheric concentration as well as the duration of their seasonal expression, mold spores seem more likely than pollens to be principals in any abatement of ILIs, including COVID-19. Because certain mold spores, e.g., Aspergillus, can propagate in man if left unattended by innate immune effectors, it also follows mold spores should be prioritized over pollens. Nonviral bioaerosols could abate viral activity by either direct or indirect means. By direct means, they might produce substances that limit viral propagation, or they might complex with viruses, limiting viral infectivity [42,43]. However, if direct antiviral activity is an attribute of the bioaerosols themselves, then one would not expect, a priori, significant disparity between individual susceptibilities to severe flu or COVID-19 [44,45]. As indirect means, others have proposed pollens stimulate the human immune system in such a way as to either potentiate endogenous antiviral activity or elicit a protective allergic response [3]. Against these proposals, asthma does not confer protection against either influenza or COVID-19 [46][47][48].
Given these, one can imagine the engagement of TLR4 by aerosols of all sorts and microscopic/submicroscopic sizes, including, but not limited to, viruses (diam 0.01-0.30 µm), mold spores (diam 1-50 µm) and pollens (diam 10-1000 µm), in a fashion analogous to the engagement of hook-and-loop adhesives, i.e., Velcro ® . Instead of loops, however, spinous processes of the various aerosols engage TLR4 'hooks,' effecting an innate immune response, the nature of which depends on the arrangement and density of the engagement. The large surface area of the extracellular domain of TLR4, 6000-8500 Å 2 ensures accommodation of many such protuberances which, in turn, explains the broad specificity of the receptor [65]. In addition, just as hook-and-loop adhesives can be rendered nonfunctional/dysfunctional by nonspecific adherence of extraneous materials, so too might TLR4 hooks become saturated with one ligand to the exclusion of another. As for the role of fibrin(ogen) D-domains, their overexpression as endogenous ligands may represent an attempt by the innate immune system to purge/disengage TLR4 from pathogenic aerosols of all sorts for, perhaps, restorative purpose.
The data presented herein bring new appreciation and understanding to seasonality and suggest a remarkable interplay between bioaerosols that influence the health of man. Indeed, considering that humans have co-existed with plants, fungi and viruses for some time, it stands to reason that, over the course of evolution, the respiratory system of the former would have developed means to cope with the significant recurring, i.e., annual, inhalational exposure to reproductive elements of the latter. As the environment-facing interface of the respiratory tree, epithelial cells and their entourage of innate immune effectors seem ideally positioned to provide that coping mechanism. That being the case, nebulized materials that exploit competition either between the various bioaerosols or between the bioaerosols and endogenous TLR4 ligands, e.g., C-terminus of the fibrinogen γ-chain [66][67][68], might prove therapeutic. Finally, and notwithstanding allergic potential, the indoor cultivation-including mold-rich fertilization-of pollenating plants might be exploited to limit occurrence of sharply seasonal ILIs.